Tune Kibana 4.4 to run on a t2 micro AWS EC2

Hi all,

I've been operating my Kibana on the same node as my first ES node for a while now. So the first node of ES is running all 3 components of ELK on the same machine.

Recently I wanted to split off Kibana and Logstash onto their own nodes. So that ES would not have to share it's memory with any other applications.

I fired up 3 t2 micros on AWS EC2 and installed Kibana there. I configured it and had all 3 nodes working with my ES setup. They were running in 3 different accounts, so I could stay within the free tier for the first year.

So the interface was identical to the one I had working on the first ES node. After letting them work for a while, if everything was still working, I'd put them behind a load balancer and have some redundancy.

But what I've found after running the 3 t2 micro's all night was that they would get overwhelmed, run out of memory and become inaccessible. Meaning that I couldn't ssh into them, and the kibana web interfaces that I'd seen previously were not available. It'd just be a grey screen.

I'd have to log into AWS, shut down the nodes one by one, and then start them back up again. Then they'd be fine. And then they would run for a while again, before becoming inaccessible again.

So what I'd like to know is if it's possible to tune Kibana to run within the memory available on a t2 micro instance. Which is 1GB.

So in short, is there a way I can get Kibana to run withing 1GB of memory so that I can stay within the free tier on Amazon.

I hadn't gotten around to trying this with Logstash yet, by splitting off Logstash into it's own node(s). I wanted to see if I could get this to happen with Kibana first.

There's an issue with recent Kibana releases chewing up all memory - related to node, I believe.

Bummer. Ok, so if I run a slightly earlier version of kibana on my aws nodes (t2), I should be good? Running it in only 348MB vm's is impressive. What version would you recommend I go down to?

Is it safe to say that you guys are working on the memorey consumption issue in the latest kibana's? :slightly_smiling:

Hey @jaddison!! Thanks I was able to get this to work using the thread you linked to!

I just added

NODE_OPTIONS="--max-old-space-size=200"

To the top of my bin/kibana fle and where the aws t2 micro I was hosting it on would crash after running all night, now has been running continuously for a couple of days!! That absolutly did the trick.

However the kibana team should really incorporate that into an upcoming release rather than relying on the user to find out how to fix this problem and adding it into the kibana script themselves.

Thanks for the link!!

Tim